Magnetometer calibration based on improved least squares and trust domain Dogleg method
Aiming at the poor accuracy of the traditional least square(LS)ellipsoid fitting method and the axi-al misalignment of magnetometer and accelerometer when the sampling points are not abundant in the process of three-axis magnetometer calibration,this paper proposes a calibration method combining improved LS with trust domain Dogleg method.Firstly,according to the error model of magnetometer,the improved LS algorithm is used to fit the ellipsoid.Then,the mathematical model of misalignment error is constructed by using the property of dot product constancy between gravity vector and geomagnetic vector.Finally,the trust domain Dogleg algorithm is used to estimate these error parameters.Simulation and experimental results are shown as below:When the im-proved LS method is used to estimate these error parameters,the error range reaches the order of 10-6.The root-mean-square error of the magnetic vector modulus is reduced by 88.5%compared with LS,which effectively improves the accuracy of the traditional fitting method.After axial alignment,the measurement accuracy of final heading angle is within±0.8°,and the root-mean-square error is less than 0.5°,significantly smaller than the error before the alignment,which proves the effectiveness of the trust domain Dogleg algorithm.
国家科技期刊平台
NSTL
万方数据
